| 1 |
On Sat, 31 May 2014 22:42:17 +0100 |
| 2 |
Fabio Erculiani <lxnay@g.o> wrote: |
| 3 |
|
| 4 |
> On Sat, May 31, 2014 at 9:06 PM, Robin H. Johnson |
| 5 |
> <robbat2@g.o> wrote: |
| 6 |
> > No, I don't agree that kernel configs "belong" to kernel packages. |
| 7 |
> > In general, barring the crazy option explosion, these are meant to |
| 8 |
> > be stock working configs that should in combination with ANY kernel |
| 9 |
> > package, produce a working kernel. |
| 10 |
> > |
| 11 |
> |
| 12 |
> Then you are just moving the problem around. |
| 13 |
> I believe that kernel configs should be provided by their own kernel |
| 14 |
> packages (and there are some, not just gentoo-sources) because it is |
| 15 |
> much easier to keep them in sync on every new release and deal with |
| 16 |
> each version separately if/as needed (including testing!). How are you |
| 17 |
> dealing with config var name changes between different kernel versions |
| 18 |
> or just different pkgs then? |
| 19 |
|
| 20 |
Different packages is not a problem; since the difference in terms of |
| 21 |
config between separate packages is small enough, a dozen of options. |
| 22 |
|
| 23 |
Different version may be a problem, a rather small one; nothing |
| 24 |
prevents one from keeping config options around for both versions. |
| 25 |
|
| 26 |
> You cannot possibly support all kernel versions for all kernel pkgs |
| 27 |
> available in tree with just one single config file in a sane, clean |
| 28 |
> and maintainable way, hoping that a change in this file will not |
| 29 |
> affect previous or future kernel releases. How are you going to test |
| 30 |
> your config changes against old kernel pkgs? Each test is quite |
| 31 |
> expensive to run. |
| 32 |
> |
| 33 |
> Good luck with that :-) |
| 34 |
|
| 35 |
Does it really need to be sane, clean and maintainable for it to work? |
| 36 |
|
| 37 |
Yes, maybe; but how sane, clean and maintainable? We can do better... |
| 38 |
|
| 39 |
A fork (eg. hardened-sources config, geek-sources config) where needed |
| 40 |
might still be a way out; however, one should consider to look into a |
| 41 |
better architecture than plain forks. For example, a config that |
| 42 |
sources a generic config and adds hardened changes on top of that; |
| 43 |
kind of like the way GRUB 2's /etc/grub.d/ config generation works. |
| 44 |
|
| 45 |
We should introduce this only when needed to avoid to over-design it. |
| 46 |
|
| 47 |
-- |
| 48 |
With kind regards, |
| 49 |
|
| 50 |
Tom Wijsman (TomWij) |
| 51 |
Gentoo Developer |
| 52 |
|
| 53 |
E-mail address : TomWij@g.o |
| 54 |
GPG Public Key : 6D34E57D |
| 55 |
GPG Fingerprint : C165 AF18 AB4C 400B C3D2 ABF0 95B2 1FCD 6D34 E57D |
Archives